multiple cargo, multiple depot reload?

[ymp-vtl]

I was referring to https://github.com/google/or-tools/blob/master/ortools/constraint_solver/samples/cvrp_reload.py for the multi depot visit in order to reset capacity to 0.
Problem i have is that i am carrying multiple cargo types in my vehicle, and each vehicle also has different capacity for each cargo type. how should i model the resetting of the capacity to 0 in this case?

for example, my current code without the depot revisit is:
(i have 5 vehicle and 4 locations to visit as a simplified example)

    data['no_of_cargo_types'] = 3
    data['demands_cargo_0'] = [0, 1, 4, 7, 10]  # for depot + nodes 1 to 4
    data['demands_cargo_1'] = [0, 2, 5, 8, 11]
    data['demands_cargo_2'] = [0, 3, 6, 9, 12]

    # vehicle capacities for each cargo type
    data['vehicle_capacities_cargo_0'] = [10, 40, 70, 100, 130]  # capacities for vehicles 1 to 5
    data['vehicle_capacities_ cargo1'] = [20, 50, 80, 110, 140]  
    data['vehicle_capacities_cargo2'] = [30, 60, 90, 120, 150] 

In the github example, it there is only 1 capacity for 1 cargo type so the demands just takes the negative of the capacity.

    data['demands'] = \
          [0, # depot
           -_capacity, # unload depot_first
           -_capacity, # unload depot_second
           -_capacity, # unload depot_third
           -_capacity, # unload depot_fourth
           -_capacity, # unload depot_fifth
           3, 3, # 1, 2
           3, 4, # 3, 4
          ]

[Mizux]

can't you create 3 demands dimensions and 3dim*5 reload stations ?

e.g. with 2 reload station per cargo dimension aka 3dim * 2 reload = 6 reload

 data['demands_0'] = \
          [0, # depot
           -_capacity, # unload depot_1
           -_capacity, # unload depot_2
           0, # unload depot_3
           0, # unload depot_4
           0, # unload depot_5
           0, # unload depot_6
           1,
           4,
           7,
           10]
data['demands_1'] = \
          [0, # depot
           0, # unload depot_1
           0, # unload depot_2
           -_capacity, # unload depot_3
           -_capacity, # unload depot_4
           0, # unload depot_5
           0, # unload depot_6
           2,
           5,
           8,
           11]
data['demands_2'] = \
          [0, # depot
           0, # unload depot_1
           0, # unload depot_2
           0, # unload depot_3
           0, # unlooad depot_4
           -_capacity, # unload depot_5
           -_capacity, # unload depot_6
           3,
           6,
           9,
           12]

then you have SlackVar on dimension_0 for depot_1 and depot_2, SlackVar on dimension_1 for depot_3 and depot_4 etc...

[ymp-vtl]

Hi @Mizux, Thank you for your reply, I might be confused but my question pertains more to modelling the capacity when returning to the depot. In the examples I looked thru, capacity was just fixed at -15, but my scenario requires a variable capacity based on the vehicle+cargo type

[Mizux]

IIRC:

1. CumulVar at each node must be positive i.e. CumulVar(i).value() >= 0
2. CumulVar(j) = CumulVar(i) + Transit(i, j) + SlackVar(j) so providing your Maximum SlackVar value can fully cover a negative Transit to make the sum at least equal to zero is OK.

So in you scenario instead of using -15 I would use the negated value of maximum vehicle capacity among the fleet and set the SlackVar's range upper bound to this value as well and it should be ok for the solver...

[ymp-vtl]

I'll give this a try, but i have a follow up question: How can we model vehicle to vehicle transfer? the same way?

For example, vehicle A is returning back to the depot and has capacity for 5 units of cargoA and 10 units of cargoB more. Its is also in the vicinity of vehicle B. vehicle B wants to consider optionally transferring its cargo to vehicle A so that it frees up some capacity, but it also requires time to do so. i was thinking of modelling it as additional fake depots (but it's moving all the time/not always available?) and the capacity is variable based on vehicle A. would setting it to the maximum vehicle capacity also make sense?